Skip to content

Conversation

@leejet
Copy link
Owner

@leejet leejet commented Dec 16, 2025

No description provided.

@ddpasa
Copy link

ddpasa commented Dec 16, 2025

This improves #1092 . Text encoders are still not completely loaded, but it looks like the transformer is fine after this.

@leejet
Copy link
Owner Author

leejet commented Dec 17, 2025

@ddpasa This update should make the LoRA for text_encoders work. However, I couldn’t find a corresponding LoRA to test it with. If you could upload a LoRA to HuggingFace or Civitai for me to download and test, I would really appreciate it.

@ddpasa
Copy link

ddpasa commented Dec 17, 2025

@ddpasa This update should make the LoRA for text_encoders work. However, I couldn’t find a corresponding LoRA to test it with. If you could upload a LoRA to HuggingFace or Civitai for me to download and test, I would really appreciate it.

Thanks @leejet , I uploaded an example Lora here: https://huggingface.co/ddhfacc/lora_debug/tree/main It has T5 and Clip layers as well.

Pulling the additional commit did not reduce the number of unused lora layers for me.

@leejet
Copy link
Owner Author

leejet commented Dec 17, 2025

@ddpasa Everything should be ok now.

@ddpasa
Copy link

ddpasa commented Dec 17, 2025

@ddpasa Everything should be ok now.

I think so, testing now. I don't see similar errors in the log after 6bff0d9. I'll generate a few and compare with Comfy.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants